Dendritic cell targeted Ccl3- and Xcl1-fusion DNA vaccines differ in induced immune responses and optimal delivery site

Fusing antigens to chemokines to target antigen presenting cells (APC) is a promising method for enhancing immunogenicity of DNA vaccines. However, it is unclear how different chemokines compare in terms of immune potentiating effects. Here we compare Ccl3- and Xcl1-fusion vaccines containing hemagglutinin (HA) from influenza A delivered by intramuscular (i.m.) or intradermal (i.d.) DNA vaccination. Xcl1 fusion vaccines target cDC1s, and enhance proliferation of CD4+ and CD8+ T cells in vitro. In contrast, Ccl3 target both cDC1 and cDC2, but only enhance CD4+ T cell proliferation in combination with cDC2. When Ccl3- or Xcl1-HA fusion vaccines were administered by i.m. DNA immunization, both vaccines induced Th1-polarized immune responses with antibodies of the IgG2a/IgG2b subclass and IFNγ-secreting T cells. After i.d. DNA vaccination, however, only Xcl1-HA maintained a Th1 polarized response and induced even higher numbers of IFNγ-secreting T cells. Consequently, Xcl1-HA induced superior protection against influenza infection compared to Ccl3-HA after i.d. immunization. Interestingly, i.m. immunization with Ccl3-HA induced the strongest overall in vivo cytotoxicity, despite not inducing OT-I proliferation in vitro. In summary, our results highlight important differences between Ccl3- and Xcl1- targeted DNA vaccines suggesting that chemokine fusion vaccines can be tailor-made for different diseases.

that induce different downstream signaling events, which may have adjuvant effects. Adding to the complexity of DNA vaccination and electroporation, the site of injection can also influence the resulting immune responses. The most common delivery routes of DNA +EP are intramuscular (i.m) or intradermal (i.d) injections. In terms of administration, i.d. immunization +EP is easier to administrate and is generally considered less painful compared to i.m. delivery +EP. However, studies have suggested that i.m. DNA vaccination, with or without EP, is more efficient at inducing cellular immune responses in both mice and macaques 23,24 .
In this study we compare the use of Xcl1 and Ccl3-fusion DNA vaccines for the induction of immune responses against influenza A hemagglutinin (HA). Xcl1 is a ligand of the receptor Xcr1, which is selectively expressed on conventional dendritic cell type 1 (cDC1) 25,26 . In contrast, Ccl3 acts as a ligand for the receptors Ccr1, Ccr3 and Ccr5 that are expressed on a number of different cell types including cDC1, cDC2, pDC, Langerhans cells, macrophages, NK cells and CD4 + and CD8 + T cells [27][28][29][30][31][32] . Consequently, Ccl3-fusion vaccines have a relatively broad tropism and target a number of different cell types that can act as APCs.
We here observe that DNA immunization with Ccl3-or Xcl1-fusion vaccines induce qualitatively different immune responses, and that the site of immunization influences the resulting immune responses. i.m. DNA immunization was most efficient when using Ccl3-fusion vaccines, and resulted in strong cytotoxic T cell responses. In contrast, Xcl1-fusion vaccines induced stronger responses after i.d. DNA immunization, which also correlated with better protection against influenza infection. In summary, our result suggest that different chemokine fusion vaccines and delivery routes may be suitable for different vaccination purposes, depending on the desired type of immune response.

Characterization of Xcl1-and Ccl3-targeted fusion vaccines. To compare immune responses against
Xcl1-and Ccl3-fusion DNA vaccines, we utilized a dimeric vaccine molecule format (Vaccibody) previously constructed in our laboratory 33 . The vaccibody consists of a targeting unit (either Xcl1 or Ccl3), a dimerization unit consisting of the C H 3 and a shortened hinge region from human IgG3 and an antigenic unit ( Supplementary  Fig. S1A). While fusing antigen to Xcl1 is an efficient method for selectively delivering antigen to Xcr1 + cDC1 6,34,35 , the surface receptors for Ccl3 are more promiscuously expressed on APC (reviewed in 31 ). To evaluate how efficiently Ccl3 and Xcl1 target cDC, we generated and purified Ccl3-mCherry and Xcl1-mCherry fusion molecules and evaluated staining of bone marrow derived DCs (BM DC) 21 . Anti-NIP-mCherry (containing a scFv specific for the hapten 5-iodo-4-hydroxy-3-nitrophenacetyl (referred to as NIP)) was included as a non-targeted control 36 . While Xcl1-mCherry predominantly stained BM cDC1 (defined as CD11c + CD45R − CD24 high ), Ccl3-mCherry stained both cDC1 and cDC2 (CD11c + CD45R − CD11b high ) to a similar degree (Fig. 1A, Supplementary  Fig. S1B). There was some staining of cDC2 with Xcl1-mCherry compared to the NIP-mCherry, but this may be a results of glycosaminoglycan (GAG) binding properties of Xcl1 37 . Consistent with this interpretation, Xcl1-mCherry selectively induced chemotaxis of cDC1 in a transwell chemotaxis assay, while Ccl3-mCherry induced a similar degree of chemotaxis of both cDC1 and cDC2 (Fig. 1B). To investigate if Xcl1 or Ccl3 could directly activate cDCs, BM DCs were incubated with 0.5 μg anti-NIP-, Ccl3-or Xcl1-mCherry for 18 h and the expression of CD40, CD80 and CD86 evaluated by flow cytometry. Neither Ccl3-nor Xcl1-mCherry induced detectable activation of cDC1 or cDC2 as defined by upregulation of CD40, CD80 or CD86, which is in accordance with previous publications for Xcl1 35 (Supplementary Fig. S1C).
To ensure targeting of cDC in vivo under similar conditions, Ccl3-, Xcl1-or anti-NIP-mCherry were injected i.v. into BALB/c mice and spleens harvested after 2 hours. cDCs and macrophages were gated as recently published ( Supplementary Fig. S1D) 38 , and evaluated for mCherry staining. As observed in vitro, Xcl1-mCherry delivered i.v. specifically stained splenic cDC1, while Ccl3-mCherry stained cDC1 and cDC2 to a similar degree (Fig. 1C). For Ccl3-mCherry we also observed a slightly enhanced staining of macrophages (Fig. 1C). In summary, these results demonstrate that Xcl1-fusion vaccines selectively target cDC1, while Ccl3-fusion vaccines target both cDC1 and cDC2.
To evaluate the ability of Xcl1-and Ccl3-fusion vaccines to induce proliferation of T cells, cell trace violet (CTV) labelled OT-I and OT-II cells were incubated with sorted BM cDC1 or cDC2 in the presence of Xcl1-, Ccl3-or NIP-OVA for 4 days. Proliferation was determined by CTV dye dilution by flow cytometry (Supplementary Fig. S2A). Xcl1-and Ccl3-OVA induced significantly higher proliferation of OT-II cells compared NIP-OVA when incubated with BM cDC1 (Fig. 1D). Interestingly, Xcl1-OVA also induced higher proliferation of OT-II cells compared to Ccl3-OVA when incubated with BM cDC1. In contrast, Ccl3-OVA induce higher proliferation of OT-II cells compared to NIP-OVA when incubated with BM cDC2 (Fig. 1E). There was an increase in proliferation of OT-II seen with Xcl1-OVA and cDC2, but the difference was not significant compared to NIP-OVA. Surprisingly, only Xcl1-OVA induced proliferation of OT-I cells when incubated with cDC1 ( Fig. 1F). No OT-I proliferation was seen with cDC2 incubated with either Xcl1-or Ccl3-OVA, although we did observe proliferation when the cells were incubated with the OVA derived peptide SINFEEKL as a positive control ( Supplementary Fig. 2B,C).

T cell responses after intramuscular or intradermal DNA vaccination.
To compare immune responses induced by Ccl3-and Xcl1-fusion vaccines, the major surface antigen Hemagglutinin (HA) from influenza A/34/PR/8 (PR8) was used as an antigen 5,6,39 . Xcl1-HA and Ccl3-HA constructs were expressed at similar levels in vitro as determined by ELISA on supernatants from transiently transfected HEK293E cells ( Supplementary Fig. S3A). The sizes of the expressed vaccibodies under reducing and non-reducing conditions were analyzed by SDS-PAGE, and confirmed that the vaccibodies were predominantly secreted as dimers ( Supplementary Fig. S3B).
Immune responses induced by Xcl1-HA and Ccl3-HA DNA vaccines were evaluated in BALB/c mice immunized by either i.m. or i.d. administration of plasmids encoding the fusion vaccines. To enhance uptake  of DNA and subsequent immune responses, the injection site was electroporated by delivering short electric pulses using either an Elgen 40 (i.m.) or a DermaVax 41 (i.d.) delivery system. T cell responses were evaluated in spleens of BALB/C mice 2 weeks after a single immunization. The number of IFNγ-secreting cells were analyzed by ELISPOT after stimulation with a MHC-I restricted peptide (IYSTVASSL) or a MHC-II restricted peptide (HNTNGVTAACSHEG), as indications of CD8 + and CD4 + T cell responses, respectively. i.d. DNA immunization with Xcl1-HA induced significantly higher numbers of IFNγ-secreting CD8 + T cells compared to Ccl3-HA ( Fig. 2A). In contrast, i.m. delivery resulted in higher number of IFNγ-secreting CD8 + T cells in CCL3-HA immunized mice compared to Xcl1-HA, although the difference did not reach significance. i.m. immunization with Ccl3-HA did, however, induce significantly higher numbers of IFNγ-secreting CD8 + T cells compared to i.d. immunization with Ccl3-HA ( Fig. 2A). No significant differences were observed in the number of IFNγ-secreting CD4 + T cells between Xcl1-HA and Ccl3-HA immunized mice after either i.d. or i.m. delivery, although there was a tendency for Xcl1-HA to induce higher numbers after i.d. immunization ( Fig. 2A). Indeed, i.d. immunization with Xcl1-HA induced significantly more of IFNγ-secreting CD4 + T cells compared to i.m. immunization with Xcl1-HA ( Fig. 2A).
To test for effector functions of the induced T cells, we performed an in vivo cytotoxicity assay. BALB/c mice were DNA vaccinated once by i.d. or i.m. immunization and injected 2 weeks later with cell trace violet (CTV) labeled splenocytes pulsed with the IYSTVASSL peptide (or a control peptide). Specific killing of the IYSTVASSL-pulsed splenocytes was analyzed after 18 hours in spleens. Surprisingly, mice immunized with Ccl3-HA displayed higher cytotoxicity compared to Xcl1-HA after both i.d. and i.m. immunization, although the difference was only significant after i.m. delivery (Fig. 3B). This observation is in contrast to the in vitro proliferation assay and the i.d. DNA immunization where Xcl1-fusion vaccines induced stronger CD8 + T cell responses compared to Ccl3-fusion vaccines. There was a tendency for Xcl1-HA to induce higher cytotoxicity after i.d. immunization, and for Ccl3-HA to induce higher cytotoxicity after i.m. immunization, but the differences did not reach significance (Fig. 3B). cDC1s have been reported to be superior at cross-presenting antigen to CD8 + T cells 42 . Since Ccl3-OVA failed to induce OT-I proliferation when incubated with cDC1 in vitro we wanted to test if Ccl3-HA induced cytotoxic T cell responses in the absence of cDC1. We therefore repeated the cytotoxicity assay after i.m. immunization with Xcl1-HA or Ccl3-HA in BATF3 −/− mice that lack cDC1 43 . As expected, Xcl1-HA immunization did not induce cytotoxicity in the absence of cDC1 (Fig. 3C). More surprisingly, i.m. Ccl3-HA immunization in the BATF3 −/− mice also failed to induce cytotoxicity, suggesting that cDC1 are equally important for cytotoxicity when immunizing with Ccl3-fusion vaccines.

Antibody responses after intramuscular or intradermal DNA immunization.
To evaluate induction of antibodies, serum samples were harvested 2 weeks after i.m. or i.d. DNA immunization and analyzed for the presence of HA-specific antibodies of the IgG1, IgG2a and IgG2b subclasses ( Fig. 3A-C). Following i.m. DNA immunization, no significant differences were observed between Ccl3-HA and Xcl1-HA for any of the three IgG subclasses, although Xcl1-HA displayed a tendency to induce higher titers of IgG2b ( Fig. 3A-C). There was also no difference for the IgG2a/IgG1 ratio, with both Ccl3-HA and Xcl1-HA inducing a predominantly Th1 polarized antibody responses (Fig. 3D). As seen for the T cell responses, i.m. immunization with Ccl3-HA induced significantly higher titers of IgG2a and IgG2b compared to i.d. immunization with Ccl3-HA. Xcl1-HA however, displayed no significant differences in antibody responses after i.d. and i.m. immunization.
Similar to i.m. DNA immunization, BALB/c mice immunized by i.d. DNA vaccination displayed no difference in the induction of HA specific IgG1 between Xcl1-and Ccl3-HA (Fig. 3A). However, Xcl1-HA induced significantly higher responses of HA specific IgG2a and IgG2b after i.d. DNA immunization compared to Ccl3-HA (Fig. 2B,C), which correlates well with the higher IFNγ secreting CD4 + T cell responses. Consequently, mice i.d. immunized with Xcl1-HA displayed a significantly higher IgG2a/IgG1 ratio compared to Ccl3-HA, suggesting a more Th1-polarized immune response (Fig. 3D). i.m. immunization with Ccl3-HA also induced higher IgG2a/ IgG1 ratio compared to i.d. delivery of the same vaccine (Fig. 3D).
Intradermal DNA vaccination with Xcl1-HA induces superior protection against a high dose of influenza virus. The above results suggest that both the chemokine and the site of DNA immunization influences the immune responses. To test the protective efficacy of the targeted vaccines, we utilized an infection  (Fig. 4A,B). In contrast, BALB/c mice immunized with NaCl succumbed to the infection by day 8 (mice were euthanized if they lost more than 20% of their starting weight as a human endpoint).
To better differentiate the efficacy of Ccl3-HA or Xcl1-HA, BALB/c mice were challenged with a high dose of influenza A PR8 virus (50xLD50) two weeks after a single intramuscular or intradermal DNA immunization ( Fig. 4C-F). As expected, challenging with a higher pathogen burden resulted in increased weight loss in the immunized groups, suggestive of increased morbidity. After intramuscular immunization, however, there were no difference in morbidity between BALB/c mice immunized with Xcl1-HA or Ccl3-HA (Fig. 4C), with both groups losing weight until day 5-6 and then recovering from infection. Consequently, the overall survival was similar between the two groups with 3/11 (27%) and 4/12 (33%) mice in the Ccl3-HA and Xcl1-HA groups having to be euthanized, respectively (Fig. 4D).
In contrast, BALB/c mice receiving a single i.d. vaccination with Xcl1-HA encoding plasmids displayed significantly lower morbidity compared to Ccl3-HA immunized mice when challenged with 50xLD50 PR8 virus after two weeks (Fig. 4E). Indeed, all 12 mice immunized with Xcl1-HA survived a high challenge dose, while 4/12 (33%) Ccl3-HA immunized mice had to be euthanized (Fig. 4F). Although the overall survival was similar, Ccl3-HA-immunized mice displayed significantly lower morbidity after i.m. immunization compared to i.d. immunization ( Supplementary Fig. 4A,B). In contrast, i.d. DNA immunization with Xcl1-HA induced significantly lower morbidity and mortality compared to i.m. immunization with Xcl1-HA ( Supplementary Fig. 4C,D).
In summary, both Ccl3-and Xcl1-fusion vaccines enhanced induction of T cell responses when given as DNA vaccines. However, the site of immunization should be taken into consideration since Ccl3 fusion vaccines generally induced more Th1 polarized antibody response with stronger cytotoxic T cell responses after i.m. immunization, while Xcl1 fusion vaccines more efficiently induced T cell responses and improved protection in an influenza mode after i.d. immunization.

Discussion
Here we present a comparison of Ccl3-HA and Xcl1-HA DNA vaccines delivered by i.d. or i.m. injections in combination with electroporation. Our observations indicate that immunization with Xcl1-HA induced stronger cellular and antibody responses when delivered by i.d. injection, which resulted in better protection against influenza infection. In contrast, Ccl3-HA induced stronger immune responses after i.m. immunization, and induced higher cytotoxic T cell responses compared to Xcl1-HA.
There are few studies comparing i.d. and i.m. delivery of DNA vaccines in combination with electroporation. Indeed, most studies do not includes electroporation of the injection sites 23,44,45 , or only includes electroporation for one of the delivery methods 46 . Nevertheless, delivery of DNA by i.m. injection has been reported to be a more efficient method for inducing Th1 associated immune responses such as IgG2a antibodies 23,44 , in addition to inducing stronger cytotoxic T cell responses 24,44 . In contrast, delivery of DNA i.d. has been reported to induce a more Th2/IgG1 polarized response 23,44 , and more protective antibody responses 46 . Our observations with Ccl3-HA fusion vaccines correlate well with these observations, with i.m. immunization inducing stronger T cell responses and a more Th1 polarized antibody response compared to i.d. immunization with Ccl3-HA. In contrast, DNA immunization with Xcl1-HA induced a similarly Th1 polarized immune response independent of i.m. or i.d. delivery. There was a tendency for higher IgG2a/IgG1 ratio after i.m. immunization with Xcl1-HA, but the difference to i.d. immunization was not significant. Indeed, i.d. immunization with Xcl1-HA induced higher numbers of IFNγ-secreting T cells compared to i.m. immunization. Consequently, our observations indicate that Xcl1-fusion vaccines are capable of enhancing Th1 polarization independent of immunization site. Interestingly, both Xcl1 and Ccl3 have been reported to be Th1 associated chemokines that enhance Th1 polarization in combination with IFNγ 47 . However, our observations suggest that Xcl1 is more potent at promoting Th1 polarization compared to Ccl3.
Surprisingly, Ccl3-HA induced stronger cytotoxic T cell responses independent of immunization site compared to Xcl1-HA, although the difference was only significant after i.m. delivery. Considering that Xcr1 is selectively expressed on cDC1 that excel at cross-presenting antigen to CD8 + T cell 25,26,42 , this observations was contrary to expectations. In addition, in vitro proliferation of OT-I cells was also only observed with Xcl1-OVA, and not when Ccl3-OVA was incubated with cDC1. However, Ccl3-mCherry was able to target cDC2s which have been reported to efficiently cross-present antigen to CD8 + T cells under inflammatory conditions 48,49 . Nevertheless, the observations in BATF3 −/− mice, suggest that Ccl3-HA was dependent on cDC1 for induction of cytotoxicity, even when delivering DNA with electroporation that induces inflammation 50 . Ccl3 has been reported to be important for activation of CD11c + CD11b + CD8α − DCs after viral infection 51 , although in our experiments, Ccl3-mCherry did not activate BM derived cDC1 or cDC2 after in vitro incubation. It is possible that the enhanced cytotoxicity seen with Ccl3-HA is mediated indirectly through a second cell population or additional signaling molecules. Studies by Brewitz and colleagues have indicated that pDCs can enhance CD8 T cell responses induced by cDC1, and that recruitment of CCR5 + pDC to CD8 + T cell-cDC1 complexes can be mediated via Ccl3, which could explain our observations 32 .
Although Ccl3-HA induced the strongest cellular cytotoxicity, Xcl1-HA provided better protection against a high dose challenge with influenza A virus. While it is likely that HA specific antibody responses contribute to this protection, we have previously observed that the protection mediated by i.d. immunization with Xcl1-HA is largely dependent on CD8 + T cells 6 . It is possible that targeting using Ccl3-and Xcl1-fusion vaccines differentially influence T cell tissue imprinting and subsequent migration. cDC1 have been reported to be important for the generation of tissue resident CD8 + T cell 52 , which are thought to play an important role in mediating protection against influenza hetero-subtypic infection 53 .
In conclusion, our observations suggest that DNA vaccination with plasmids encoding Xcl1-and Ccl3-fusion molecules differs in their optimal site of delivery. Since Ccl3 and Xcl1 target receptors that are differentially expressed on cDC1 and cDC2, it is possible that the frequency of different DC populations in the tissue influence Scientific RepoRts | (2019) 9:1820 | https://doi.org/10.1038/s41598-018-38080-7 the resulting immune response. Under steady state condition, however, the frequency of cDC2 has been reported to be considerably higher than cDC1 in both skin and muscle 48,54 . Electroporation has been reported to induce a massive influx of DCs, in addition to other APCs 40 , but it is unclear if this effect influences the relative frequency of cDC1 and cDC2. Of relevance, we have previously observed that delivery of Xcl1-OVA fusion protein via laserporation of the skin induce cytotoxic T cell responses and reduce tumor growth in a B16 melanoma model 34 .
Together with the data presented here, these observations suggest that skin is an efficient delivery site for targeting cDC1. Chemokine fusion vaccines are currently being tested in clinical trials, and in summary, our observations suggest that both targeting strategy and delivery site should be considered depending on the desired type of immune responses. Generation and purification of targeted vaccibodies. The construction of the Xcl1-targeted and CCL3-targeted vaccibodies have been described previously 5,6 . Purification of NIP-, Xcl1-and CCL3-vaccibodies with mCherry or OVA as antigen were performed as described in Gudjonsson et al. 21 . In brief, HEK293E cell were transiently transfected in 5-layer BD Multi-Flasks (Corning) using Lipofectamine 2000 (Invitrogen). Supernatants were harvested after 3-4 days and applied to a column containing CaptureSelect FcXL affinity Matrix (Life Technologies) connected to an ÄKTAprime plus chromatography system (GE healthcare). Bound vaccibodies were washed with PBS and eluted using 0.1 M glycin-HCl pH 2.7. Eluted proteins were dialyzed twice against PBS, and subsequently concentrated using Vivaspin20, 50.000 MWCO cutoff columns (Sartorius).

Generation of bone marrow derived DCs (BM DCs).
The protocol for generating BM DCs by incubation with Flt3L was originally published by Brasel and colleagues 55 , and was described in detail in our previous publication 21 . For evaluation of targeting and activation after incubation with Xcl1-or CCL3-mCherry, BM derived DCs were harvested 9 days after incubation with 0.1 μg/ml recombinant murine Flt3L and seeded at a density of 4 × 10 5 cells/well in a 96-well tissue culture plate. BM DCs were subsequently incubated with 0.5 μg/ml of of Xcl1-, CCL3-or NIP-mCherry and harvested after 18 h and evaluated for mCherry staining or expression of CD40, CD80 or CD86 by flow cytometry.
Chemotaxis assay. The protocol for chemotaxis assay on BM DCs was described in detail in Gudjonsson et al. 21 .
In vivo binding of Xcl1-and CCL3-mCherry. For in vivo staining, 25 μg of Xcl1-, CCL3-or NIP-mCherry were injected i.v. into BALB/c mice and the spleens harvested after 2 h. Single cell suspesions were generated using a Gentlemacs dissociator (Miltenyi), and subsequently treated with Tris-Buffered Ammonium Chloride (ACT) lysis buffer and filtered through a 70 μm Nylon strainer. To efficiently differentiate cDCs and macrophages, the single cell suspensions were stained with a combination of antibodies as described by Guilliams and colleagues 38 .
OT-I/OT-II proliferation. Bone marrow derived Flt3L DCs were identified as cDC1s and cDC2s -CD45 − CD11c + CD11b − CD24 + and CD45 − CD11c + CD11b + CD24 − , respectively, and sorted on FACSAria IIIu (BD Biosciences, Franklin Lakes, NJ). Post-sorting purity checks showed cDC1 and cDC2 populations of >99% purity. OT-I and OT-II cells were harvested from spleens of OT-I 56 and OT-II 57 TCR transgenic mice, and enriched with mouse CD8 + or CD4 + T cell magnetic bead isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany), respectively. OT-I and OT-II cells were stained with 5 μM CellTrace Violet (CTV, Molecular Probes, Eugene, OR) before being incubated with sorted DCs at a DC:T cell ratio of 1:10 for OT-I cells and 1:3 for OT-II cells. All incubations were done with 1 μg/ml of protein for 4 days at 37 °C 5% CO 2 , and finally the T cells were analyzed for proliferation by flow cytometry. As a positive control, cells were incubated with 1 μg/ml OVA 257-264 or OVA 323-339 peptide for OT-I or OT-II, respectively.

ELISA.
ELISAs were performed as previously described 6  IFNγ ELISPOT. Single cell suspensions of splenocytes were prepared using the GentleMACS dissociator according to the manufacturers enzyme-free protocol. To detect IFNγ secreted by splenocytes, an ELISpotPLUS for Mouse IFNγ kit with pre-coated anti-IFNγ plates was used in accordance with the manufacturers protocol (MABTECH AB, Nack Straand, Sweden). In short, spleens were dissociated, treated with Tris-Buffered Ammonium Chloride (ACT) lysis buffer and filtered through a 70 μm Nylon strainer to prepare single cell suspensions. Cells were added to the plates at a concentration of 0.5 × 10 6 and restimulated with the HA-derived peptides IYSTVASSL (MHC-I, H-2K d restricted) or HNTNGVTAACSHEG (MHC-II, I-E d -restricted), or negative control peptide at a concentration of 2 μg/ml for 18 hours at 37 °C 5% CO 2 . Plates were automatically counted and analyzed using a CTL ELISPOT reader (CTL Europe GmbH, Bonn, Germany). Values obtained from negative control peptide wells were subtracted from values obtained from stimulation with specific peptides for each sample.
Cytotoxicity assay. In vivo cytotoxicity assay were modified from Durward et al. 58 . In brief, splenocytes were harvested and incubated with 1 μg/ml of the influenza HA peptide IYSTVASSL or a control peptide at a density of 5 × 10 7 cells/ml for 1 h at 4 °C before being transferred to 37 °C for an additional 30 min. Peptide-loaded cells were washed twice in PBS and labeled with either 1 μM (negative control) or 10 μM (IYSTVASSL) cell trace violet (CTV) at a density of 5 × 10 7 cells/ml for 20 min at 37 °C. After washing twice in PBS, the cells were re-suspended in PBS at a density 5 × 10 7 cells/ml and the two populations mixed 1:1. A total of 1 × 10 7 mixed cells were injected i.v. into BALB/c mice immunized 14 days earlier with Xcl1-HA or CCL3-HA by either intramuscular or intradermal DNA vaccination. Spleens were harvested 18 h later, and processed into single cell suspension as described for IFNγ ELISPOT. The ratio of CTV low to CTV high cells were determined by flow cytometric analysis, and cytotoxicity calculated as % specific lysis = [1 − (non-transferred control ratio/experimental ratio)] × 100.
Statistics. All statistics were calculated using Prism 6.0 software (GraphPad Software, La Jolla, CA). For all figures where multiple statistical comparisons were performed on data in the same figure, such as in vitro and in vivo binding, antibody titers, ELISPOT assays and cytotox assay, two-tailed one-way ANOVA was performed with Tukey's or Dunn's multiple comparison test. For chemotaxis data, a two-tailed t-test was performed. For analysis of weight curves, two-way ANOVAs were performed. For survival curves Mantel-Cox test was performed.

Data Availability
Materials, data and protocols will be made available to readers upon request to the corresponding authors.